GLUTAMATE TRANSPORTERS AROUND THE TRIPARTITE
SYNAPSE
by
Silvia Holmseth
Centre for Molecular Biology and Neuroscience
Department of Anatomy, Institute of Basic Medical Sciences, University of Oslo
2011
© Silvia Holmseth, 2011
Series of dissertations submitted to the Faculty of Medicine, University of Oslo No. 1239
ISBN 978-82-8264-287-3
All rights reserved. No part of this publication may be
reproduced or transmitted, in any form or by any means, without permission.
Cover: Inger Sandved Anfinsen.
Printed in Norway: AIT Oslo AS.
Produced in co-operation with Unipub.
The thesis is produced by Unipub merely in connection with the
Table of contents
Acknowledgements………5
Abbreviations……….6
Minireview ……….7
Aims ………..20
Methods……….21
Synopsis……….28
- Summary………28
- Original papers………..29
- Comments on individual papers………...30
Conclusions……….. 38
References……….40
Papers………53
Acknowledgements
The present work was performed at the Department of Anatomy, Institute of Basic Medical Sciences, University of Oslo and the Centre for Molecular Biology and Neuroscience (CMBN), Norway. I am grateful for the great facilities being provided to me.
I would like to thank my principal supervisor Niels Christian Danbolt and my co-
supervisor Knut Petter Lehre for excellent guidance and for always taking their time. You have been great motivators and mentors ever since I was introduced to science when starting my master in 2001.
I am grateful to all my colleagues and collaborators who contributed to the papers in this thesis. Thanks to past and present co-workers in the Neurotransporter Group. Also special thanks for technical assistance to Henriette Danbolt. Also thanks to the directors of the department of Anatomy during these years. Finally, I would like to thank my closest family and friends for backing med up and believing in me.
The work has been funded by The Norwegian Advanced Research Program (Toppforskningsprogrammet).
Silvia Holmseth Oslo, September 2011
Abbreviations
ALS Amyotrophic lateral sclerosis
AMPA -amino-3-hydroxy-5-methyl-4-isoxazole propionic acid ASCT Neutral amino acid exchanger
CNS Central nervous system DHK Dihydrokainate
EAAT Excitatory amino acid transporter (= glutamate transporter) EAAC1 Excitatory amino acid carrier (EAAT3)
GAD Glutamic acid decarboxylase GDH Glutamate dehydrogenase
GLAST Glutamate aspartate transporter (EAAT1) GLT1 Glutamate transporter (EAAT2)
GSH Glutathione
GTRAP Glutamate transporter associated protein mGLUR Metabotropic glutamate receptors MSO Methionine sulfoximine NAC Sodium dicarboxylate transporter NMDA N-methyl-D-aspartate
PAG Phosphate activated glutaminase PSD Postsynaptic density
SLC Superfamily of solute carriers TBOA threo--benzyloxyaspartate VGLUT Vesicular glutamate transporter xCT Glutamate-cystine exchanger
Minireview
In addition to being an amino acid and a component of proteins, glutamate is the main excitatory neurotransmitter in the central nervous system (CNS) (for review see: Fonnum, 1984; Ottersen and Storm-Mathisen, 1984; Danbolt, 2001). Glutamate is involved in most aspects of normal brain function including cognition, memory and learning. Brain tissue contains large amounts of glutamate, around 5-15 mmol per kg depending on the region (Schousboe, 1981). The extracellular concentrations are kept low and are in the order of a few micromolar (Hamberger et al., 1983), or may be even lower (Herman and Jahr, 2007). The highest glutamate concentrations are found intracellularly in glia cells, nerve terminals and synaptic vesicles (in increasing order) (Ottersen et al., 1992). It is suggested a concentration of more than 60 mM inside synaptic vesicles (Shupliakov et al., 1992). The concentration in cytosol is not known, but assumed to be in the low millimolar range implying that the concentration gradient across the plasma membrane is several thousand fold.
Figure 1: The concentration gradients across the plasma membranes are great. The extracellular glutamate concentration is around 1 M, inside nerve terminals it is around 10 mM and 100 mM in synaptic vesicles (for references see: Danbolt, 2001).
When glutamate has been released into the extracellular space, it binds to glutamate receptors on neuronal and glial cell membranes to exert its signaling role. The glutamate receptors are divided into three families (for review see: Kristensen et al., 2006). One
family is the metabotropic receptors which are coupled to G-proteins (mGluR1-8).
Activation of these receptors leads to changes in inositol phosphate or cyclic nucleotide metabolism. The two other families are ionotropic receptors which mean that they are glutamate gated ion channels that conduct Na+ or Ca2+. They are named after the glutamate analogues that activate them. One of them is NMDA (N-methyl-D-aspartate) receptors which has high affinity for glutamate and are slowly inactivating. The last family is AMPA (-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid) or kainate receptors according to their preference for AMPA or kainate. AMPA/kainate receptors are rapidly inactivating and have low affinity for glutamate.
Because glutamate modulates fundamental neurological processes and because glutamate only can act on the glutamate receptors from the outside, the extracellular concentrations must be tightly controlled. Both too much and too little receptor
stimulation can be harmful l(Danbolt, 2001). There is no enzyme extracellularly that can metabolize glutamate, so the glutamate must be removed by cellular uptake (for review see: Danbolt, 2001). Although simple diffusion may be important for the reduction of glutamate in the synaptic cleft in the submillisecond timescale (Clements, 1996), diffusion can only cause glutamate redistribution. It cannot cause real removal from the extracellular fluid. The only mechanism for net removal is cellular uptake mediated by the glutamate transporters located in both neurons and astroglia (for review see: Danbolt, 2001).
The close proximity of astrocytes to synapses points towards glia as a part of the synaptic functional unit. The term “the tripartite synapse” has been introduced in recognition of the contribution of astrocytes to synaptic function (Volterra et al., 2002).
In addition to clearance of neurotransmitters, it has been proposed that astrocytes can have neuron–like activities like releasing glutamate (Bezzi and Volterra, 2001). This, however, is still debated. For instance, a recent transcriptome database does not give support to the notion that astrocytes express the proteins involved in vesicular glutamate release (Cahoy et al., 2008) and it remains to be shown if glutamate containing vesicles exists in astrocytes.
Glutamate transporters
Glutamate transporters are also called “sodium and potassium coupled glutamate transporters” or “excitatory amino acid transporters” (“EAATs”). The term “sodium dependent high-affinity transporters” have earlier been in use, but because the affinity not is particular high (Km varying between 1-100 M depending on the subtype and assay method) and because the transporters also depend on potassium, this term is no longer in common use (Danbolt, 2001). Five distinct glutamate transporter subtypes have been cloned from animal and human tissue and have been assigned to the solute carrier family (slc) number 1: GLAST (EAAT1; slc1a3; Storck et al., 1992), GLT1 (EAAT2; slc1a2;
Pines et al., 1992), EAAC1 (EAAT3; slc1a1; Kanai and Hediger, 1992), EAAT4 (slc1a6;
Fairman et al., 1995), and EAAT5 (slc1a7; Arriza et al., 1997). In addition several splice- variants of the various EAATs have been reported (Utsunomiya-Tate et al., 1997; Meyer et al., 1998a; Münch et al., 1998; Meyer et al., 1999; Huggett et al., 2000; Rauen et al., 2004; Rozyczka and Engele, 2005). Two transporters for neutral amino acids (alanine serine cysteine transporter; ASCT1 and 2; gene slc1a4 and 5) have also been assigned to this family. There are also other transporters capable of transporting glutamate. For instance, in mitochondria there are “mitochondrial glutamate transporters” (for review see: Sluse, 1996) and in synaptic vesicles (Varoqui et al., 2002; Chaudhry et al., 2008) there are “vesicular glutamate transporters” (VGLUT1-3; slc17a6-8). In the plasma membrane there are transporters for neutral amino acids (e.g. ASCT2; slc1a5) and dicarboxylates (e.g.NAC3; gene slc13a5) that can transport glutamate with low affinity.
In addition there is a glutamate-cystine exchanger (xCT; slc7a11) that exchange extracellular cystine with intracellular glutamate (Sontheimer, 2003; Sato et al., 2005).
Transport mechanism and stoichiometry
Glutamate transporters use the electrochemical gradients of Na+, K+ and H+ to transport glutamate into the cells (Kanner and Sharon, 1978). In that way they are secondary active transporters because the gradients for Na+ and K+ are maintained by the Na+/ K+-ATPase which requires ATP to work. The transport is electrogenic because more positive charge moves in than out in each transport cycle. The transport is therefore stimulated by the negative membrane potential. The glutamate transporters can work in both directions and
are described as shuttles that transport either glutamate, Na+ and H+ or K+. When glutamate/ Na+/ H+ is transported into the cell, either K+ or glutamate/ Na+/ H+ have to be transported out before starting a new transport cycle. Na+ is required for glutamate binding and K+ is required for net transport. In the absence of K+, the transporters are locked in exchange mode and can only exchange external substrate with internal substrate (Danbolt, 2001). It is important to note that failed cycles occur in between completed (net) transport cycles (see Volterra et al., 1996). Exchange can be considered incomplete or as failed transport cycles. The stochiometry (the fixed number of ions involved in the transport) of the transporters have been determined to be 1 glutamate, 3 Na+, 1 H+, 1 K+ for EAAT1-3 (Zerangue and Kavanaugh, 1996a: Levy et al., 1998; Owe et al., 2006).
This is an important parameter as it determines the concentrative capacity of the transporters, their energy consumption and the sensitivity to changes in ion gradients. In addition to the stoichometric transport the transporters also function as ion channels (in particular EAAT4 and 5, but almost non-existing in GLT1. For review see Seal and Amara, 1999) and as water channels (at least in the case of EAAT1) (MacAulay et al., 2001).
Figure 2: The stoichiometric transport of glutamate transporters. The transporters operate as shuttles that catalyze the stoichiometric transport of 1 glutamate, 3 sodium and 1 hydrogen ion
inwards in exchange with 1 potassium ion. In addition 400 water molecules are transported to the inside and the transporters can operate as ion channels.
Substrate selectivity
The glutamate transporter’s specificity for glutamate is not absolute, because they can transport L-glutamate, D-aspartate, L-aspartate and L-cysteate with nearly the same affinity. In addition several analogs have been made. The glutamate molecule is very flexible and can have many different conformations. This explains how transporters, receptors and enzymes can bind glutamate despite having very different binding sites.
This is what medicinal chemists exploit when they construct new molecules mimicking different glutamate conformations. Transportable glutamate analogs with high affinity include threo--hydroxyaspartate (THA), L-trans-pyrrolidine-2,4-dicarboxylic acid, L- serine-O-sulfate (L-SOS) and (2S,1’S,2’R)-2-(2-carboxycyclopropyl)glycine (CCG-III).
There are also some non-transportable analogs, blockers, like dihydrokainic acid (DHK) and threo--benzyloxyaspartate (TBOA) derivatives (for review see: Bridges et al., 1999;
Shimamoto and Shigeri, 2006).
Glutamate transporter structure
EAAT1-5 are glycoproteins consisting of between 500 and 600 amino acids. Their molecular masses are in the range of 60-75 kDa. They share 50-60 % amino acid sequence. Parts of the sequence are highly conserved between the EAATs, suggesting these parts play an important role. For instance, residues 396-400 (GLT1 nomenclature) are shown to be important for transport activity and to participate in sodium binding (Zarbiv et al., 1998; Zhang et al., 1998). And the residues 403-404 appear to be involved in potassium binding (Kavanaugh et al., 1997). Based on the crystal structure of a bacterial glutamate transporter homologue (Yernool et al., 2004), the transporters are predicted to have 8 transmembrane domains with intracellular amino- and carboxy- terminals. It is bowl-shaped with a large aqueous basin which makes glutamate easily reach binding sites halfway into the membrane, this architecture is well suited for rapid binding of glutamate in synapses. It has a triangular shape with sides of 80 Å. They most
likely both exist and work as trimers (Haugeto et al., 1996; Koch and Larsson, 2005;
Grewer and Rauen, 2005).
Localizations and quantifications
The five different EAATs are found in different amounts in different areas of the brain.
Quantification in absolute terms is required to know which glutamate transporter subtypes are dominating in different brain regions. The absolute amounts are also needed to determine their capacity in mathematical models (e.g. Rusakov and Kullmann, 1998) and to answer whether there is enough of the protein for its proposed role.
Immunohistochemical studies have suggested that GLAST and GLT1 are both expressed in the plasma membranes of astrocytes facing neuropil (Lehre et al., 1995; Chaudhry et al., 1995; Lehre and Danbolt, 1998; Furness and Lehre, 1997; Furuta et al., 1997a; Furuta et al., 1997b; Furuta et al., 1997a Shibata et al., 1997). It is also shown that GLAST and GLT1 coexist in the same astrocytic membrane (Lehre et al., 1995; Haugeto et al., 1996), but as different homooloigomeric complexes (Haugeto et al., 1996). GLAST is the major glutamate transporter in the cerebellum (Lehre et al., 1995), the inner ear (Furness and Lehre, 1997; Takumi et al., 1997), the circumvenrtricular organs Berger and Hediger, 2000) and in the retina (Derouiche and Rauen, 1995), while GLT1 is the quantitative dominating glutamate transporter in the forebrain: about 1% of total brain protein is GLT1 (Danbolt et al., 1992; Lehre and Danbolt, 1998). GLAST and GLT1 are expressed at low levels at birth, but increases dramatically during development (Ullensvang et al., 1997). The highest increase is during synaptogenesis (P14-P28). The concentrations of GLAST and GLT1 are very high: 15 000 and 23 000 molecules per m3 in the stratum radiatum of hippocampus and molecular layer of cerebellum, respectively (see also Table 1 below). By employing a stereological method to estimate the cell membrane area containing these transporters, it was possible to calculate the density of transporters in the membranes (Lehre and Danbolt, 1998).
EAAT4 is mainly found in the Purkinje cell spines and dendrites in the cerebellar molecular layer (Dehnes et al., 1998). It is expressed at low levels in the forebrain. In the cerebellum the amount of EAAT4 is 0.2 mg/g tissue giving a density of 1800 molecules per m2 spinemembrane.
EAAT5 is only found expressed in retina (Arriza et al., 1997), where it is localized in Müller cells and neurons (Eliasof et al., 1998). It seems that the associated chloride conductance is more important for EAAT5’s physiological role than the transport function (Veruki et al., 2006; for review see Wadiche and von Gersdorff, 2006).
EAAC1 is a neuronal glutamate transporter (see paper IV included in this thesis).
GLT1 splice variants
The mRNA encoding GLT1 is large (11 kb; Pines et al., 1992) and has been found to have more than 30 splice variants; some of which translate into variant proteins (Utsunomiya-Tate et al., 1997; Meyer et al., 1998b; Münch et al., 1998; Meyer et al., 1999; Rozyczka and Engele, 2005; Rauen et al., 2004). Alternatively splicing of both N- and C-terminal exist. The functional properties are not changed by altering the termini (Sullivan et al., 2004). There exist at least three different C-terminal GLT1-variants (Rauen et al., 2004). The first studies on GLT1 protein localization (Danbolt et al., 1992;
Lehre et al., 1993; Levy et al., 1993; Chaudhry et al., 1995; Lehre et al., 1995) used antibodies recognizing all isoforms. Subsequent antibodies (Rothstein et al., 1994;
Schmitt et al., 1996) raised to the extreme C-terminus of GLT1a gave a seemingly identical labeling pattern. In contrast, the first reports on the distribution of GLT1b (protein and mRNA) in the brain were conflicting as some investigators detected it in neurons (Schmitt et al., 2002; Chen et al., 2002; Kugler and Schmitt, 2003; Reagan et al., 2004), including nerve terminals (Chen et al., 2002), while others only observed
astroglial expression of the protein (Reye et al., 2002; Sullivan et al., 2004). Chen and co- workers later concluded that their anti-GLT1b antibodies were not specific (Chen et al., 2004). Instead they found that antibodies to GLT1a labeled (in hippocampus CA1) a subset of axon terminals and spines in addition to astroglia. Berger and co-workers (2005) demonstrate that the probes used for detection of GLT1b mRNA have an unspecific component, and conclude that both isoforms are widely expressed in astrocytes and that GLT1a is the predominant neuronal isoform. The expression of GLT1a protein in terminals was confirmed in the hippocampus (Furness et al., 2008) and in the somatic sensory cortex (Melone et al., 2009). Four different GLT1 N-terminal
variants also exist (Utsunomiya-Tate et al., 1997; Rozyczka and Engele, 2005), although it is unknown which of N-terminal variants that are expressed in the CNS in vivo.
Glutamate is recycled (the glutamate-glutamine cycle)
From the described localization and quantification data above it follows that the majority of glutamate uptake is catalyzed by GLT1 which is mainly localized in glia (Paper V). In order to be recycled, this means that the glutamate must be further transported to get back to the nerve terminals. In glia glutamate is detoxified to glutamine by glutamine
synthetase. Glutamine is transported out of the glia cell by glutamine transporters and taken up by nerve terminals by another glutamine transporter. Inside mitochondria of the nerve terminal glutamine is reconverted to glutamate by the neuron-specific phosphate activated glutaminase (PAG). The glutamate is then loaded into vesicles by vesicular glutamate transporters (VGLUTs) and ready for use again.
Figure 3: Recycling of glutamate by the glutamate-glutamine cycle. Glutamate is released from the nerve terminal and taken up by glutamate transporters in the dendritic spine (I), the nerve terminal (VI) and in the glial cell (II). Inside the glial cell it is detoxified to glutamine by glutamine synthetase, released to the extracellular space by a glutamine transporter (III) and taken up in the neuron by another glutamine transporter (IV). (Reproduced from Danbolt 2001).
The importance of the glutamate-glutamine-cycle is discussed, but there are several evidences for its importance. Inhibition of glutamine synthetase by methionine sulfoximine (MSO) leads to a loss of synaptic activity in the retina and depletion of neuronal glutamate (Laake et al., 1995; Barnett and Pow, 2000). Inhibition of glutamine synthetase also inhibits epileptiform activity on hippocampal cultures (Bacci et al., 2002).
Also the majority of the glutamate transporters are localized in the proximity of synapses and glutamine synthetase is exclusively a glial enzyme while PAG is neuronal. On the other hand, de novo synthesis of glutamate from glucose has considerable function because each glutamate molecule, on the average, only can be recycled 3-4 times before being oxidatively degraded by the Krebs cycle (Hertz et al., 1999). Glutamatergic neurons can also sustain glutamate release independent of glutamine by pyruvate carboxylation (Hassel and Bråthe, 2000), although it has earlier been reported that pyruvate carboxylase is an astrocytic enzyme (Yu et al., 1983).
Uptake of glutamate into nerve terminals
It seems bothersome to do this recycling through glial cells. The most efficient would be direct uptake of glutamate into the nerve terminals. On the other hand, the uptake is electrogenic contributing to depolarization, and this may be a reason why evolution has favored sharing the burden with astrocytes. This uptake has been a mystery. Several studies have shown that this uptake exists, but the protein responsible has unvaded detection (see comments to Paper V).
Regulation
Glutamate uptake can be regulated on virtually all levels: DNA transcription, mRNA splicing, protein synthesis, targeting, glutamate transport and associated ion channel activity (Danbolt, 2001). A number of substances and proteins have shown to affect uptake activity and expression of glutamate transporters. Interestingly, several regulatory mechanisms can have differential effect on different subtypes of glutamate transporters.
Dependence of neurons for glial glutamate transporter expression is also shown in vivo when GLAST and GLT1 levels are reduced in striatum after lesioning of afferent
glutamatergic nerve fibers (Levy et al., 1995; Ginsberg et al., 1995). When astrocytes are cultured in the presence of neurons they express both GLT1 and GLAST, while pure astroglial cultures only express GLAST (Gegelashvili et al., 1997a; Swanson et al., 1997). Glutamate uptake is thought to be modulated by it’s substrate glutamate. When glutamate binds to glutamate receptors, it is thought that this may represent a feedback regulatory mechanism for glutamate uptake. Long-term treatment of astrocytes in culture with L-glutamate (0.1-3 mM) resulted in a dose-dependent increase in uptake activity (Gegelashvili et al., 1996). In addition glutamate uptake is modulated by fatty acids like arachidonic acid (Trotti et al., 1995; Zerangue et al., 1995; Dunlop et al., 1999), by phosphorylation (for review see Gonzalez and Robinson, 2004), by red-ox mechanisms (Volterra et al., 1994; Trotti et al., 1997), and by growth factors, cytokines and soluble factors (Gegelashvili et al., 1996; Gegelashvili et al., 1997b; Schlag et al., 1998).
Glutamate transporters are kept in position and modulated by proteins connected to the cytoskeleton. Four such proteins have so far been characterized: Ajuba which binds GLT1 (Marie et al., 2002), GTRAP3-18 which binds EAAC1 (Lin et al., 2001), and GTRAP41 and GTRAP48 that binds EAAT4 (Jackson et al., 2001). Increased GTRAP3- 18 expression in cells reduces glutamate transport by lowering substrate affinity of EAAC1. This is also seen in another study where glutathione is decreased when GTRAP3-18 expression is increased, then GTRAP3-18 modulate EAAC1s uptake of cysteine which is necessary to make glutathione (Watabe et al., 2008). GTRAP3-18 might act as an endogenous inhibitor of EAAC1. On the other side Ajuba is not found to alter glutamate transport mediated by GLT1, Ajuba's function with respect to glutamate transport is unclear. NMDA-receptor subunits also interact with EAAC1 to control surface expression, suggesting a close relationship between glutamate receptors and transporters (Waxman et al., 2007). In general, the machinery for EAAC1-trafficking seems to be more developed than for the other transporters. Which is logical since EAAC1 is expressed in both kidneys and intestine (Paper IV) where metabolic needs rapidly changes. Further evidence of differential regulation of glutamate transporter subtype expression is seen during development (Ullensvang et al., 1997, Furuta et al., 1997a). Rothstein et al. (2005) reported that -lactam antibiotics (e.g. ceftriaxone)
selectively elevate GLT1 expression. Ceftriaxone appeared to be cytoprotective both in vitro and in vivo in a mouse model of ALS (Rothstein et al., 2005).
The importance of glutamate transporters
As shown (Lehre and Danbolt, 1998) the concentration of the major glutamate
transporters are very high, 15 000 and 23 000 molecules per m3 in the stratum radiatum of hippocampus and molecular layer of cerebellum, respectively. Whether the glutamate transporters manage to clear the synaptic released glutamate will depend on several factors. First, it depends on how much glutamate is released. One m3 contains about one glutamatergic nerve terminal (Woolley and McEwen, 1992). It is assumed that one vesicle is released at the time, and that each vesicle contain some 400-5000 glutamate molecules (Clements, 1996), which is a broad range. Second, the morphology of the synapse matters. Most of the glutamate transporters are located on astrocytes. If the transporters shall contribute to clear the glutamate, they must be located near the glutamate release sites. Because localization of astrocytes in relation to synapses varies between regions and different synapses, the role of glutamate transporters will also vary.
Third, the number of glutamate transporters will be important. The high number of glutamate transporters are sufficiently high to clear the number of glutamate molecules released per vesicle. Since the glutamate transporters are slow (about 70 ms per cycle (Wadiche et al., 1995; Auger and Attwell, 2000), they compensate by their high number and high affinity that easily make them “buffer” the extracellular glutamate. GLT1 clears most of the glutamate with its high number and localization in astrocytic membranes close to synapses, GLAST and EAAC1 seem to play minor roles. It is likely that glutamate transporters serve at least dual functions both as a transporter of other substances than glutamate and as an ion-channel (Ryan et al., 2004; Veruki et al., 2006).
Why there are five different glutamate transporters and several splice variants that are so differentially expressed and regulated is a question.
The physiological roles of glutamate transporters
It is obvious that glutamate transporters play a significant role in removing glutamate from the extracellular space. The abundance and importance of GLT1 is also evident from the phenotype of GLT1 deficient mice. In mice deficient in GLT1 glutamate levels
rise enough to cause epilepsy and cell death (Tanaka, 1997). Similarly, the GLAST knockout mouse show reduced motor coordination and have major changes in the retina (Watase et al., 1998), which is in agreement with the localization of GLAST. The EAAC1 knockout do not develop remarkable neurological symptoms, except for reduced spontaneous locomotor activity and dicarboxylic aminoaciduria (Peghini et al., 1997).
EAAC1 also transports cysteine with an affinity 10–20-fold higher than that of GLAST or GLT-1 (Zerangue and Kavanaugh, 1996b). A recent study demonstrated age- dependent neurodegeneration with decreased glutathione (GSH) content, increased oxidant levels and increased susceptibility to oxidative stress in EAAC1-deficient mice (Aoyama et al., 2006). These EAAC1-deficient mice also showed an age-dependent decrease in neuronal number in the substantia nigra (Aoyama et al., 2008a). GSH plays an important role in detoxifying reactive oxygen species (ROS) and thereby protecting cells from oxidative stress (for review see Dringen, 2000). GSH is a tripeptide composed by cysteine, glycine and glutamate where cysteine is the rate-limiting factor for GSH- synthesis in neurons. In primary neuron culture, approximately 90% of total cysteine uptake is mediated by sodium-dependent systems, mainly excitatory amino acid transporters (EAATs) (Chen and Swanson, 2003; Himi et al., 2003).
Glutamate transporters and disease
Malfunction of glutamate uptake has been reported in many diseases, but it is not easy to differentiate between cause and effect of dysfunction of glutamate transporters (for review, see Danbolt, 2001;Beart and O'Shea, 2007). The role of glutamate transporters in disease are often connected to glutamate as a neurotoxin. At high extracellular
concentrations glutamate is toxic to the brain and can contribute to neuronal cell death (Choi, 1992). If the extracellular glutamate is increased more glutamate will bind to glutamate receptors. Binding to the ionotropic glutamate receptors will give more glutamate release and an influx of Na+ and Ca2+. The cells energy consumption will increase in order to pump these ions out again. In addition free radical production will increase (Bondy and Lee, 1993). This may in turn impair energy production and glutamate uptake, and maybe even reverse glutamate transporters causing further
glutamate release. An increase in extracellular glutamate concentration easily starts vicious circles.
In ischemia, hypoxia and hypoglycemia, ion gradients dissipate, glutamate transporters reverse and a massive efflux of glutamate and cell death is the result (Nicholls and Attwell, 1990; Rossi et al., 2000). In addition there is a vesicular release (Drejer et al., 1985) and release through swelling-activated anion channels (Kimelberg and Mongin, 1998). It is still debated from which compartments most of the glutamate leak during ischemia (Ottersen et al., 1996; Lipton, 1999). The most favored hypothesis is that the glutamate comes from neurons because most brain glutamate is stored there.
Further, the glutamate uptake in glia cells are less sensitive to hypoxia than the neuronal uptake (Swanson et al., 1994). Quantitative immunocytochemistry studying the changes in glutamate concentrations following brain ischemia strongly suggest that the release of glutamate is by neurons (Ottersen et al., 1996), and by the dendro-somatic compartments.
ALS-amyotrophic lateral sclerosis is characterized by muscular weakness, atrophy and spacticity (Chancellor and Warlow, 1992). It is caused by death of motor neurons where exitotoxicity is one possible reason. A dramatic loss of GLT1 in ALS- patients have been shown (Rothstein et al., 1995), but another paper has shown no change (Milton et al., 1997). Whether reduced GLT1-expression is a primary event leading to motor neuron death or if it is secondary to the neuron death is not clear.
Also in Alzheimer's disease there are several contradictory findings about reduced levels of GLT1 (Li et al., 1997) while another report found no correlation between reduced GLT1 levels and Alzheimer (Beckstrøm et al., 1999). It has been suggested that oxidation is a cause of disturbed GLT1 expression and that GLT1 splice variants occur in several neurological diseases like Alzheimer and ALS (Honig et al., 2000).
It has been observed elevated levels of glutamate in patients with various epilepsy (e.g. Janjua, 1992). And mice deficient of GLT1 develops spontaneous seizures (Tanaka, 1997). In contrast to the mentioned neurological disorders it is questioned if it is too much glutamate transport in schizophrenia (for review, see: Carlsson et al., 1997). A glutamatergic hypofunction can be due to hyperactive glutamate transporters, defective receptors, lack of receptors, inadequate glutamate release or lack of glutamatergic nerve terminals.
AIMS
The overall aim of this thesis has been to increase our insight into glutamatergic transmission and reuptake by providing quantitative data on the distribution and expression levels of the various glutamate transporter subtypes. The thesis has specially searched to give answer to these questions:
-What is the functional significance of EAAC1?
-Which transporter is responsible for the uptake of glutamate into nerve terminals?
- How abundant are the various C-terminal variants of GLT1?
Methods
I will here give an overview of the main methods I have used in this study focusing on principles without repeating all the details as they are described in the papers.
Tissue preservation-fixation
The purpose of fixation is to preserve the tissue components of interest after death and during the cutting and labeling procedures. It is important to realize the type of fixative and the type of fixation procedure have to be adapted to the purpose. The fixation should preserve the structures without impairing detection. Because fixation is usually only the beginning of a long process, and it is important to consider the entire procedure before fixing the tissue.
Fixatives: These are chemicals that chemically modify the tissue in such way that the desired preservation is obtained. The most common is formaldehyde which binds to and make covalent bonds between adjacent amino groups. Formaldehyde has a tendency to polymerize, and the polymer diffuses slower into the tissue. To get an efficient and quick fixation it is important to work with monomers. This is usually accomplished by depolymerizing paraformaldehyde to formaldehyde shortly before use. Glutaraldehyde works in the same way, but is a slightly larger molecule which diffuse slower, but being a di-aldehyde it forms cross-links more efficiently. This is an advantage with respect to immobilizing diffusible molecules and with respect to preservation of ultrastructure, but not when it comes to antigen accessibility. Note that fixation time, concentration and temperature are just as important parameters as the choice of chemical compounds.
Fixation procedure: Small tissue pieces can be efficiently fixed simply by immersing them in fixative. The best fixation of thicker tissue (like a whole brain) is obtained by a perfusion of the animal through the heart. The animal is given a lethal dose of pentobarbital, then a tubing with a needle is put into the left ventricle-aorta and the fixative delivered through the tubing by a peristaltic pump.
Preparation of antibodies
Antibodies are immunoglobulins, proteins which specifically recognize their antigens.
They are composed of two heavy and two light chains and are Y-shaped, the two heavy
chains are connected together by disulfide-groups. Between the light and the heavy chain there is an antigen binding site. In research antibodies are used to identify and localize proteins.
Antibodies have been made by immunizing animals with the antigen of interest. We have mostly used synthetic peptides corresponding to parts of the protein. The advantages with using short peptides are that we can avoid conserved parts and thereby avoid cross- reaction with other proteins. Further we know exactly which part of the protein it reacts to and the antibody can so be used in functional studies of the transporter proteins. The disadvantage with antibodies against short peptides is that they may not recognize the native protein, only the peptide, and it can be difficult to choose the most antigenic parts of the protein. Special computer programs exist that can help us choose the most hydrophilic and assumed most antigenic parts of the protein. But the programs only give an indication.
EAAC1
GLT1
Figure 4: Antigenicity plot of EAAC1 and GLT1. Hydrophilic parts are shown in blue and the predicted antigen parts in pink.
It can be good to avoid transmembrane parts, as well as glycosylated and lipidated parts.
Also cysteines should be avoided as they tend to make trouble during synthesis (Danbolt et al., 1998). But there are no absolute rules. From Figure 4 one can think that it should not be more difficult to produce antibodies to EAAC1 than to GLT1, but our experience and paper I shows that this is not the case.
The peptides must be coupled to a carrier as they are too small to be antigenic (molecules less than 1-5 kDa are normally not antigenic). We have used keyhole limpet hemocyanin (KLH), BSA or thyroglobulin as carrier proteins. The peptides are then coupled to the carrier by glutaraldehyde. Most often we have used rabbits for the production. They respond well and give 70-80 ml serum per month. The antibodies are affinity purified by immobilized peptide, the same as the animal was immunized with. If the antibody show unwanted reactivity, they should also be absorbed. Then the carrier-protein is coupled to agarose and aldehyde treated, and the sera is run slowly through this column.
Immunohistochemistry
Immunohistochemistry followed by microscopy is useful for studying a protein’s localization at regional, cellular and subcellular level. Multiple labeling can also be done with different fluorochromes in order to compare labeling of several proteins at the time and see whether they colocalize or not.
Briefly, mice or rats are perfusion fixed, tissue taken out and cut on vibratome. The sections are blocked and incubated with primary antibody. The immunoreagents will diffuse into the section from the surface. The antigen-antibody complexes are visualized with biotinylated anti-rabbit IgG and streptavidin-biotinylated horseradish peroxidase complex followed by diaminobenzidine (DAB) for ordinary light microscopy or fluorescent dye for confocal microscopy.
Pre-embedding peroxidase labeling: Pre-embedding means labeling of the tissue before it eventually is embedded and processed for electron microscopy. Pre-embedding
peroxidase labeling transforms the substrate diaminobenzidine into a precipitate that will be restricted by the cell membranes. This precipitate is electron dense so that the sections can also be used for electron microscopy if wanted. This is a sensitive method and has the advantage that the sections can be viewed light microscopically first. Interesting parts of the section can be cut out and embedded in a plastic material (Durcupan®) for electron microscopy. The method is not quantitative as structures may be unlabelled either because they do not contain the antigen or because the antigen is not accessible.
Post-embedding immunogold: Post-embedding labeling means that labeling is done after embedding and cutting of the sections. The antibody-antigen complex is viewed with a secondary antibody coupled to a gold particle (the gold particle is electron dense) This method is quantitative and gives higher resolution as it can show in which parts of the membrane the protein is localized, but it is less sensitive than the pre-embedding labeling.
Figure 5. Illustration of pre-embedding (left) and post-embedding (right) techniques. For pre- embedding the reaction product is restricted by the cell membrane. For post-embedding the immunoreagents are not restricted by cell membranes. The length of the antigen-antibody complex is around 20 nm, so the gold particle can be on the outside of the structure although the antigen is on the inside (reproduced from Danbolt et al., 1998).
Microscopy
For ordinary light microscopy the resolution is limited by the wavelength of the light used. The resolution (d) is decided by the formula: d=(0.61x)/NA where is the wavelength of the light and NA is the numerical aperture. Optimally the resolution can be 0.2 m, but most often it is closer to 0.5 m. This means that the light microscope is excellent for getting an overview over which part of the section and which structures that are labeled, but the subcellular labeling needs higher resolution. For instance nerve terminals which are about 500 nm will appear as dots in this microscope.
The electron microscope has high resolution and is extremely powerful for detecting the exact position of a protein at the ultrastructural level. Here the light is replaced by an electron beam, which has a wavelength of 0.005 nm. Since the resolution is proportional with it gives a very high resolution, normally around 0.2 nm
For fluorescent microscopy the antigen-antibody complex is viewed with a fluorescent dye (or particle like e.g. quantum dot) with a certain excitation and emission spectrum.
Since fluorescent dyes exist with different emission spectra it can be performed multiple labeling. For multiple labeling you have to ensure that the emission spectra do not overlap. This allows us to study colocalization of proteins. The immunofluorescence technique requires application of antibodies which must be carefully validated by appropriate control experiments.
In a conventional (i.e., wide-field) microscope the entire specimen is evenly illuminated resulting excitation of the entire section thickness. Light originating from the parts of the section that are out of focus often causes a strong background. The background from unfocused parts can be avoided by using confocal microscopy. The confocal microscopy provides sharper images than the conventional microscope since it excludes light from outside the focal plane. Each image represents a thin cross-section of the specimen. It is also possible to obtain a three-dimensional reconstruction by sampling several images along the vertical axis of the specimen. The resolution of confocal microscopy is close to the theoretical limit (see above).
Western blotting
The purpose of blotting is to transfer the molecular species from the matrix of a gel (agarose or polyacrylamide) and to the surface of a membrane (nitrocellulose, PVDF, nylon or others) where they can be studied much more easily. In general proteins in a tissue homogenate are separated by molecular weight by SDS-PAGE (Laemmli, 1970).
The proteins are subsequently transferred from a gel to a membrane The protein can then be immunolabeled with the desired antibody. Probed antigens can be visualized by a variety of secondary antibodies conjugated to e.g alkaline phosphatase or horseradish peroxidase. The immunocomplex can be detected by different methods like chemoluminecence, fluorescence or radioactivity. The method can be quantitative if specific controls or calibration systems are included. It is then important to work within the linearity of the signal. We used chemoluminecence which is a very sensitive method, the signal was captured by films or a CCD camera (a CCD camera converts optical
brightness into electrical amplitude using charge coupled device (CCD), the signals can be converted to digital values). Films are very sensitive, but the CCD camera can capture linear data over a broad range. The CCD camera gives digital recording and very short exposure times can be done. In such a way saturation during image acquisition can be avoided. It is very important to ensure precise loading of the gel and to repeat the experiments in triplicates.
Immunoblotting with purified protein extended the linear range upwards, but reduced linearity at low protein amounts presumably due to non-specific loss of protein. In paper IV this loss was prevented by adding 1 g of total brain proteins from GLT1-deficient mice to all samples containing purified GLT1 protein (thus, the amount of purified proteins loaded was varied, but the amount of unrelated brain tissue proteins was kept constant). It is also important to save pictures from different exposure times in order to analyze the linearity of the exposure times.
Immunoisolation of proteins
Immunoisolation of proteins is purifying a protein by means of immobilized antibodies.
The antibodies are first coupled to a solid phase material (like Protein-A-Sepharose Fast Flow, which is Protein A coupled to Sepharose beads). Protein A is the most known IgG binding protein. It is a Fc-receptor that means it can bind to the Fc part of the antibody, keeping the antigen binding site free to bind to the antigen. The antibody is then used to pull out the antigen from a tissue or cell extract. The antibody is covalently bound to the Protein-A-Sepharose by dimethyl suberimidate to ensure that the antibodies do not fall off in the last step when a low pH buffer is added. Brain tissue or transfected cells are homogenized in a suitable buffer. We have often used SDS as a detergent to be sure that the tissue is properly dissolved. After homogenization we have added Triton to make the solution antibody “friendly” before incubation with the immobilized antibodies.
Immobilized antibodies were incubated with the extracts for 1–2 h before the proteins were eluted with a low pH-buffer The eluted proteins are then neutralized. Protease inhibitors and a reducing agent to avoid oligomerization of the transporter proteins are added to the eluate.
Cell culturing and transfection of mammalian cells
Cell culturing is a method where cells can be kept alive ex vivo. By keeping optimal conditions the cells can grow. By transfecting the cells with foreign DNA of a protein of interest, the cells can express the proteins we want. In paper IV we wanted to verify that the antibodies to the GLT1 subtypes also could distinguish between the individual splice variants on immunoblots (as brain tissue contains a mixture of them), HEK293T (human embryonic kidney cell cultures) cells were transiently transfected with cDNA encoding GLT1 variants, solubilized in SDS, subjected to SDSPAGE and immunoblotted with the antibodies. The HEK293T-cells did not have detectable endogenous expression of GLT1, neither non-transfected cells nor cells transfected with an open vector were immunopositive The GLT1 subtypes were immunoisolated with subtype specific antibodies from the transfected cells.
SYNOPSIS
Summary
Glutamate is the major excitatory neurotransmitter in the mammalian nervous system. It is inactivated by cellular uptake catalyzed by a family of glutamate transporter proteins (GLT1, GLAST, EAAC1, EAAT4 and EAAT5). The main aim of the present thesis was to determine the contributions of the individual glutamate transporter subtypes to the total glutamate uptake around hippocampal synapses focusing on the EAAC1 subtype and on the mysterious transporter responsible for nerve terminal uptake of glutamate. The first step on this endeavor was to make antibodies to EAAC1. As outlined in Paper I, it turned out to be more difficult to make good antibodies to EAAC1 than to the other glutamate transporters. Specificity testing using tissue from EAAC1 knockout mice as negative controls revealed highly specific interactions with unrelated proteins. Paper II summarizes of the lessons learnt about immunocytochemical specificity testing, and Paper III illustrates how the antigen pre-adsorption test can be misleading. After having overcome methodological problems, we were in position to address the original question.
In Paper IV a new procedure for immunoisolation of EAAC1 was developed, and known amounts of pure EAAC1 protein was used as standard to quantify EAAC1 concentrations in brain tissue extracts. EAAC1 was found to be present at 13 g per gram hippocampal protein. This is 100 times less than GLT1 and argues against a significant contribution of EAAC1 to rapid transmitter activation. EAAC1is selectively expressed in neuronal somata and dendrites throughout the brain, and thereby in a total surface area similar to that of astrocytes. In Paper V we show that nerve terminal glutamate uptake fully depends on GLT1, and that about 10% of hippocampal GLT1 protein is expressed terminals. This also explains why high levels of GLT1 mRNA is present in CA3 pyramidal cells. In Paper VI we describe antibodies to GLT1 splice variants and show that GLT1a represents about 90 % of total hippocampal GLT1, while GLT1b and GLT1c represented 6 and 1 %, respectively. We also provide the first data on the distribution of the GLT1b and show that this variant does not contribute to nerve terminal uptake.
Original papers included in this thesis:
I) Holmseth S, Dehnes Y, Bjørnsen LP, Boulland JL, Furness DN, Bergles D, Danbolt NC (2005) Specificity of antibodies: unexpected cross-reactivity of antibodies directed against the excitatory amino acid transporter 3 (EAAT3) Neuroscience 136: 649-660.
II) Holmseth S, Lehre KP, Danbolt NC (2006) Specificity controls for immunocytochemistry. Anat Embryol (Berl) 211: 257-266.
III) Holmseth S, Zhou Y, Follin-Arbelet V, Lehre KP, Bergles D, Danbolt NC
Specificity controls for immunocytochemistry: the antigen pre-adsorption test can lead to inaccurate assessment of antibody specificity
Submitted.
IV) Holmseth S, Dehnes Y, Huang YH, Follin-Arbelet V, Grutle NJ, Mylonakou NM, Plachez C, Bergles D, Zhou Y, Furness DN, Danbolt NC and Lehre KP, (2011) Low density of EAAC1 (EAAT3; slc1a1) glutamate transporters suggests involvement in neuronal metabolism rather than in rapid control of synaptically released glutamate Manuscript.
V) Furness D, Dehnes Y, Akhtar A, Rossi D, Hamann M, Grutle N, Gundersen V, Holmseth S, Lehre K, Ullensvang K, Wojewodzic M, Zhou Y, Attwell D, Danbolt N (2008) A quantitative assessment of glutamate uptake into hippocampal synaptic terminals and astrocytes: New insights into a neuronal role for excitatory amino acid transporter 2 (GLT1). Neuroscience 157:80-94.
VI) Holmseth S, Scott HA, Real K, Lehre KP and Danbolt NC (2009)
The concentrations and distributions of three C-terminal variants of the GLT1 (GLT1;
slc1a2) glutamate transporter protein in rat brain tissue suggests differential regulation.
Neuroscience 162: 1055-71.
Comments on original papers
Comments on paper I: Specificity of antibodies: Unexpected cross-reactivity of antibodies directed against the excitatory amino acid transporter 3 (EAAC1) glutamate transporter
Background: We wanted to quantify and localize EAAC1 in order to gain insight into its physiological functions. Because we initially did not have access to EAAC1 knockout mice, we made several antibodies to different parts of the EAAC1 protein molecule in order to be as sure as possible about the specificity. Our group had already made antibodies the C-terminus (residues 491-523 and 510-523) before I got involved. These antibodies gave rise to weak labeling on Western blots, but remarkably strong labeling in tissue sections. This mismatch made us feel uneasy. Another factor that made us feel uneasy was that preferential labeling of cytoplasm. Further, there were some data in the literature based on antisense knockdown of transporters attributing about a third of the total glutamate uptake activity to EAAC1 (Rothstein et al., 1996). We then noted that other investigators (Kugler and Schmitt, 1999) had made antibodies to a peptide
corresponding to residues 480-499. They got strong labeling and described colocalization of EAAC1 and tubulin. We wondered if Rothstein and co-workers (1996) could be right after all.
Results: We made antibodies to a peptide similar to that of Kugler and Schmitt (1999), and like them, we also got antibodies that recognized a strong band. The electrophorectic mobility, however, was slightly higher than that of the band recognized by the other antibodies. Nevertheless, we got excited because this could mean that we were about to discover a novel and much more abundant variant of EAAC1. To be sure that this protein indeed was EAAC1, it was decided to try to immunoisolate it to get a protein sequence.
Then we discovered that the protein was sometimes in the water soluble fraction and sometimes in the membrane fraction. This did not make sense. Further, the antibodies labeled axons strongly. Using a robotic ELISA system, I screened the antibodies for reactivity towards a number of non-EAAC1 proteins, including actin and tubulin which
are abundant in axons. The antibodies reacted strongly with tubulin and did so with a remarkable specificity. This explained why the protein was sometimes in the water soluble fraction: it was there when the conditions favored depolymerization of tubulin! I then fractionated the antiserum and obtained one fraction that recognized both tubulin and EAAC1, and another fraction that appeared specific for EAAC1. The EAAC1- specific fraction did not label axons. The labeling appeared restricted to the
dendrosomatic compartment. There was no labeling of oligodendrocytes and no labeling of astroglia.
In parallel with this, we also made antibodies to a number of other EAAC1 peptides using different immunization protocols.
Discussion: This paper offers an explanation to why Kugler and coworkers observed a colocalization between EAAC1 and tubulin anf thereby concluded that EAAC1 is also in oligodendrocytes: their antibody probably also cross-reacted with tubulin. Unfortunately, their antibody was not available when we asked them for a sample so this hypothesis could not be tested directly. It is important to add that Kugler and co-workers only had one antibody and they did not have access to the EAAC1-knockout. This means that they did not have the means to uncover the reactivity with tubulin. Further, I had a robot to do the ELISA assays for me. During this work it became clear to us that the majority of antibodies against synthetic peptides do not recognize the native protein, and that cross- reaction with unrelated proteins is very common. Importantly, it also suggests that the pre-adsorption test has little value when testing affinity purified antibodies.
Comments on paper II: Specificity controls for immunocytochemistry
Background: Still the goal was to quantify and localize EAAC1. To be sure about the specificity of our antibodies when used to label tissue sections we got hold of fixed brains from EAAC1 knockout mouse previously described (Peghini et al., 1997).
Result: This final test of the antibodies against EAAC1gave a surprising result: despite the testing described in Paper I and despite being specific on immunoblotting, the antibodies did label something in the EAAC1 knockout. So instead of getting the data I wanted, I ended up knocking out my own paper. In stead, we summed up all the lessons learnt in this paper.
Discussion: This paper is intended as a guide for immunocytochemistry. Too many papers are published without proper controls, and it is frustrating to see this. Everybody should test their antibodies carefully before they go ahead using them. It is bad to have spent a lot of time on antibodies that are not specific. It is also bad for others that wrong results are published, and it is costly to correct erroneous data published by others. The latter point is illustrated in Paper V.
Comments on paper III: Specificity controls for immunocytochemistry: the antigen pre-adsorption test can lead to inaccurate assessment of antibody specificity
Background: A widely used test for verification of antibody specificity is the pre- adsorption test, in which the antibody is mixed with the antigen used to generate the antibody. If addition of the antigen to the antibody prior to incubation with the sample (tissue sections or Western blots) removes the ability of the antibody to label, then this is taken as proof of specificity. It is, or rather, it should be well known that this test has major limitations as pointed out before (e.g. Pool and Buijs, 1988; Swaab et al., 1977;
Burry, 2000; Holmseth et al., 2005; Holmseth et al., 2006). Nevertheless, this test continues to be used, sometimes as the only specificity test. We decided to look a bit more into this and to illustrate the point.
Results: We tested the specificity of several antibodies by using different tests; by performing the antigen pre-adsorption test, by immunoblotting, by using several antibodies to the same antigen, and by using tissue from knockout mice as negative controls. We show that antigen pre-adsorption blocks all binding of the affinity purified
antibodies for the selected antibodies shown here (antibodies to GLT1, EAAC1 and BGT1), regardless of whether this binding is to the transporters under study or to cross- reacting epitopes. Further, we demonstrate that there is no perfect correlation between specificity of labeling seen on immunoblots and labeling seen in sections. This goes both ways. Labeling in sections may be due to cross-reactivity even if the blots look perfect, and labeling in sections may be specific even if the blots are dirty. Extra bands on a blot represent a warning, but are in themselves not an absolute indicator and thereby not a sufficient argument for rejecting a study if there are other reasons to believe that the labeling is specific (e.g. no labeling in sections from a knockout mouse).
Discussion: This shows that the pre-adsorption test does not confirm the identity of the antigens being labeled. Pre-absorbing the antibodies is an alternative to affinity
purification in the sense that it shows if the labeling is due to the right antibodies, but this does not tell what the same antibodies bind to in the samples. It can be used as a test when unpurified serum is used because serum contains several antibodies. This test is so widely used that it is important to make people aware of its limitations.
Comments on paper IV: Low density of EAAC1 (EAAT3; slc1a1) glutamate transporters suggests involvement in neuronal metabolism rather than in rapid control of synaptically released glutamate
Background: In papers I-III we have described the difficulties we had with obtaining specific antibodies to EAAC1. The antibodies we had were specific on immunoblots, but when they were tested on tissue sections, they all gave labeling in the knockout. We had no antibody that was suitable for immunocytochemistry. Finally we managed to purify one antibody that was monospecific to EAAC1. This antibody was used to study EAAC1’s localization, now with the EAAC1 knockout as control.
Results: This paper consists of four main elements. The first element is the purification of EAAC1. We first tried to purify EAAC1 from brain, but found out that it was easier to
do this from kidney. After having developed a new purification protocol by modifying the procedures used with success for GLAST and GLT1, pure EAAC1 protein was eventually obtained, although in small amounts. The second main element is the determination of EAAC1 concentrations in brain tissue by immunoblotting using known amounts of pure EAAC1 protein as standards. The EAAC1 concentration in the young adult hippocampus was found to be about 100 times lower than that of GLT1. The third element was the generation of a specific EAAC1 antibody to allow localization. The strategy that finally gave rise to good antibodies was somewhat unusual: to grow anaerobic bacteria from rabbits and use extracts from these to absorb and remove unwanted antibodies from antisera prior to affinity isolation on columns with immobilized peptide. EAAC1 was found to be restricted to neuronal cell bodies and dendrites. EAAC1 was not detected in axon-terminals. Most EAAC1 is intracellular. The fourth main element is the determination of the surface area of dendrites and spines, and then to calculate the number of EAAC1 molecules using molecular mass, concentration (gm/liter) and membrane surface area.
Discussion: This paper corrects the literature in several ways. Firstly, we bring to rest the question of whether EAAC1 is expressed in glia or in terminals. Secondly, we address the question of the concentrations and show that there are not enough EAAC1 molecules to capture any significant proportion of released glutamate molecules before they can escape from the synaptic cleft even if the highest EAAC1 concentration should be, as claimed (Conti et al., 1998), in the spine membrane around the postsynaptic density. With a mean maximum density (assuming all EAAC1 is inserted in the membrane) of 90 EAAC1-molecules per m2 , and a perisynaptic zone of about 0.2 m2 it follows that there may be around 20 EAAC1 molecules per synapse (or six trimers). The number of glutamate molecules released from a synaptic vesicle is believed to be in the range 500- 5000. Even if we assume that all of the monomers can bind glutamate, it follows that EAAC1 is able to catch only about 4 % of the glutamate molecules if 500 are released and 0.4 % if 5000 are released. This makes it clear that EAAC1 is expressed at too low levels to make a significant contribution. Further, EAAC1’s ultrastructrural localization is uncertain because the post-embedding technique is not sensitive enough. It is also clear
that it is GLT1 that removes most of the synaptic glutamate. The low density of EAAC1 and that it is localized outside the synapses suggest that EAAC1 is not important for removal of extracellular glutamate, but that it may be important for neuronal metabolism because it is the only glutamate transporter expressed in the somato-dendritic
compartment of most neurons. This agrees with the claim that EAAC1 is involved in neuronal glutathione synthesis (Aoyama et al., 2008b) as EAAC1 is the primary route for cysteine uptake.
The one uncertainty with this work is whether there are splice variants of EAAC1.
We only have antibodies against the C-terminal available. So the immunoprecipitations and the following immunoblotting are both done with a C-terminal antibody. It is then possible that only a fraction of the EAAC1 protein is detected. On the other hand nobody has reported splice variants of EAAC1.
Comments on paper V: A quantitative assessment of glutamate uptake into hippocampal synaptic terminals and astrocytes: New insights into a neuronal role for excitatory amino acid transporter 2 (GLT1)
Background: It was concluded several decades ago that nerve terminals can take up glutamate (for review, see Fonnum, 1984; Ottersen and Storm-Mathisen, 1984; Danbolt, 2001). In fact, many believed that terminals had more uptake activity than glia. This conclusion, however, was weakened by the combination of the findings that GLT1 is present in astrocytes and accounts for 95 % of the total reconstitutable uptake activity (Danbolt et al., 1992), and that glial expression of GLT1 depends on intact nerve terminals (Levy et al., 1995). The presence of GLT1 in terminals was then investigated electron microscopically using antibodies to glutaraldehyde fixed D-aspartate. It was then shown that terminals do have glutamate uptake activity and that this is sodium dependent and is able to concentrate (Gundersen et al., 1993). At this time it was not known that the GLT1 can be selectively inhibited by dihydrokainate and this drug was therefore not tested. However, synaptosome preparations are inhibited by dihydrokainate (Johnston et al., 1979; Robinson et al., 1993), but it could be argued that such preparations are contaminated with glia containing GLT1. And all antibodies that had been made against
GLT1, by several different groups, labeled only glial cells in tissue sections. Another puzzling observation was the high levels of GLT mRNA in some neurons ((Torp et al., 1994; Schmitt et al., 1996; Torp et al., 1997); Berger and Hediger, 1998). Further, the GLT1 knockout mice had virtually no uptake activity (Tanaka et al., 1997). Taken together, this did not add up.
Results: We repeated the electron microscopy experiments by Gundersen and co-workers (1993) and show that the uptake into terminals is sensitive to dihydrokainate like the uptake into astrocytes. Further, we show that nerve terminals in tissue from the GLT1 knockout mice do not have the ability to accumulate D-aspartate. The uptake activity of terminals represents about half of the uptake activity in slices, and about three quarters of that in synaptosome preparations. We did not find any uptake of D-aspartate into spines, the structures that express EAAC1, that is another indication for that EAAC1 does not contribute much to removal of extracellular glutamate. We then noticed that the hippocampal slices have much larger extracellular spaces after in vitro incubation than normal perfusion fixed tissue. This meant that the various tissue components are better separated. We made use of this and did post-embedding immunogold for GLT1. Then we managed to show that terminals do express GLT1 protein. Most of the immunoreactivity is found in glial structures (about 80%), but some 5-10% is present in terminals. Some of the immunoreactivity (8%) was found in axons where it was distributed in a plasma membrane surface area several times larger than that of astroglia. This explains why CA3 pyramidal cell bodies have high levels of GLT1 mRNA.
In the mean time others (Chen et al., 2002) claimed that terminals express the b- variant of GLT1. We then showed that the a-variant is the predominant isoform in the terminals, but before we managed to publish the proof, they corrected themselves (Chen et al., 2004) explaining that their antibodies were not specific and that terminals express the a-variant.
Discussion: This work resolves the question of whether the glutamate uptake into nerve terminals is mediated by GLT1 and why CA3 pyramidal cells have GLT1 mRNA. It also support the notion that synaptosome preparations preferentially measure uptake into
terminals. However, the study left us with another question: how can 5-10 % of GLT1 molecules account for half of the uptake activity? This mismatch could be explained by damaged glial cells. Glial cells have a lot of processes and probably many of them will be cut off during the slice-preparation. This explains why the glial cells look collapsed in the slices. The nerve terminals seem to manage the slice-preparation better. They are small, round structures which seem to be well preserved during a slice-experiment. It needs to be done more studies on how the glial cells manage during these slice experiments. It is also a question whether glutamate uptake mediated by GLT1 is differentially regulated in nerve terminals versus glial cells (Xu et al., 2003; Furness et al., 2008). The capacity of the nerve terminal uptake in relation to glial uptake is not known, and we do not know how this uptake is in intact brain. To follow up glial versus nerve terminal glutamate uptake we have floxed the GLT1 gene by adding LoxP-sites to it. In combination with expression of Cre-recombinase this will delete GLT1 in the mouse. The construct works because crossing these animals with Nestin-Cre abolished GLT1 expression in the nervous system.
Comments on paper VI: The concentrations and distributions of three C-terminal variants of the GLT1 (GLT1; slc1a2) glutamate transporter protein in rat brain tissue suggests differential regulation
Background: In paper IV it was confirmed that GLT1a is expressed in nerve terminals, although at low levels. But this paper did not address the question whether other variants could be present in terminals. Quantitative data on protein levels were missing
Results: In this paper we make antibodies to three C-terminal variants of GLT1, and use these antibodies to quantify and localize the variants. To be able to obtain pure variant proteins, cells were transfected with the cDNA of the different proteins. The pure proteins were used as standards for the relative quantifications. Concentrations of total GLT1 protein were normalized using a pan-GLT1 antibody. We find that GLT1a represents about 90 % of total hippocampal GLT1, while GLT1b and GLT1c represented 6 and 1 %, respectively. We also provide the first data on the distribution of the GLT1b
on the subcellular level. The antibodies to GLT1c did give some labeling in tissue sections, but we were unable to detect differences between wild-type and knockout mice.
To document the GLT1 distributions in more detail, we also made a web-atlas for GLT1a and –b in collaboration with another group at the Centre for Molecular Biology and Neuroscience. Series throughout a whole rat and mouse brain were cut and labeled with the antibodies. The section were scanned and can be viewed as high-resolution
microscopic images showing GLT1 distributions. This is a virtual microscope in a data- repository for online inspection and re-use (http://www.rbwb.org; choose “Atlas of Neurotransporter Distributions”).
Discussion:
Here, we addressed the question of whether other GLT1 variants also are expressed in nerve terminals indirectly by comparing the amounts relative to each other. We were not able to detect GLT1b in nerve terminals. Because GLT1b is expressed at levels around 15 times lower than GLT1a, it follows that if GLT1b is in nerve terminals it must be at a level even lower than the GLT1a in terminals. It is valuable to know that the amount of the GLT1b and GLT1c is low in control tissue since it is speculated in that these variants are up-regulated in different neurological diseases. Having splice variants give
possibilities for differential regulation and targeting for the different splice variants. We here show that GLT1 splice variants seem to be differentially regulated both during development and in regional expression. The same study should also be repeated on retina since these variants are expressed there.
CONCLUSIONS
The present work demonstrates the importance of proper antibody testing before use (Paper I), and that genetically modified animals is the best negative control for immunocytochemistry (Paper II). The pre-adsorption test is not a proper specificity test as pre-adsorption with the antigen also removes eventually unspecific labeling (Paper III). We show that it is important to determine the quantities of the glutamate transporters to answer whether it is enough of the protein for its proposed functional role. The level of
EAAC1 is about 1% of the GLT1 level. This together with EAAC1’s localization outside synapses mean that EAAC1 is unlikely contribute much to synaptic clearance of
glutamate and probably plays metabolic roles (Paper IV). GLT1 is the quantitatively dominating glutamate transporter. Although this transporter is predominantly in astrocytes, it is also the transporter responsible for the uptake of glutamate into nerve terminals (Paper V). Nerve terminals express GLT1a. GLT1b is only found in astrocytes.
GLT1a represents about 90 % of total hippocampal GLT1, while GLT1b and GLT1c represented 6 and 1 %, respectively (Paper VI).
